Radiological Society of North America (RSNA) Annual Meeting

NVIDIA Presentations & Talks

Come learn how to apply AI to medical imaging and listen to talks on how deep learning can impact clinical radiology.

SUNDAY 11/25

  • 8:30 a.m. - 10:00 p.m
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This class will focus on basic concepts of convolutional neural networks (CNNs), and walk the attendee through a working example. A popular training example is the MNIST data set which consists of hand-written digits. This course will use a data set we created, that we call 'MedNIST' and consists of 1000 images each from 5 different categories: Chest X-ray, hand X-ray, Head CT, Chest CT, Abdomen CT, and Breast MRI. The task is to identify the image type. This will be used to train attendees on the basic principles and some pitfalls in training a CNN. The attendee will have the best experience if they are familiar with Python programming.
  •  
  • 10:00 a.m. - 11:30 a.m

    NVIDIA Booth #6568
    Demo

       
    •  
       
      Zebra Medical Vision develops machine learning algorithms aimed to support radiologists workflow by detecting a variety of clinical findings accurately and in a timely manner. The demo will cover the work done by Zebra to dramatically reduce time of analysis.
  •  
  • 10:30 a.m. - 12:00 p.m

    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This session will focus on preparation of the image and non-image data in order to obtain the best results from your deep learning system. It will include a discussion of different options for representing the data, how to normalize the data, particularly image data, the various options for image annotation and the benefits of each option. We will also discuss the 'after training' aspects of deep learning including validation and testing to ensure that the results are robust and reliable.
  •  
  • 11:45 a.m. - 1:15 p.m.

    NVIDIA Booth #6568
    Demo

       
    •  
       
      Imagia will demo a clinician driven end-to-end AI biomarker discovery.
  • 12:30 p.m. - 2:00 p.m.
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This class will focus on basic concepts of convolutional neural networks (CNNs), and walk the attendee through a working example. A popular training example is the MNIST data set which consists of hand-written digits. This course will use a data set we created, that we call 'MedNIST' and consists of 1000 images each from 5 different categories: Chest X-ray, hand X-ray, Head CT, Chest CT, Abdomen CT, and Breast MRI. The task is to identify the image type. This will be used to train attendees on the basic principles and some pitfalls in training a CNN. The attendee will have the best experience if they are familiar with Python programming.
  •  
  • 1:30 p.m. - 3:00 p.m.

    NVIDIA Booth #6568
    Demo

       
    •  
       
      This demo will focus on an AI solution to interpret MRI imagery with the goal of detecting conditions and quantiying them.
  •  
  • 2:30 p.m. - 4:00 p.m.
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This session will focus on the use of deep learning methods for segmentation, with particular emphasis on 3D techniques (V-Nets) applied to the challenge of MR brain segmentation. While focused on this particular problem, the concepts should generalize to other organs and image types.
  •  
  • 3:15 p.m. - 4:45 p.m.

    NVIDIA Booth #6568
    Demo

       
    •  
       
      12 Sigma Technologies will demo a deep-learning-based CAD system for lung nodule detection.

MONDAY 11/26

  • 8:30 a.m. - 10:00 p.m.
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This class will focus on basic concepts of convolutional neural networks (CNNs), and walk the attendee through a working example. A popular training example is the MNIST data set which consists of hand-written digits. This course will use a data set we created, that we call 'MedNIST' and consists of 1000 images each from 5 different categories: Chest X-ray, hand X-ray, Head CT, Chest CT, Abdomen CT, and Breast MRI. The task is to identify the image type. This will be used to train attendees on the basic principles and some pitfalls in training a CNN. The attendee will have the best experience if they are familiar with Python programming.
  •  
  • 10:00 a.m. - 11:30 p.m.
    NVIDIA Booth #6568
    Demo

       
    •  
       
      This demo will focus on the Quibim Precision platform with AI algorithms and imaging biomarkers analysis for hospitals and clinical trials.
  •  
  • 10:30 a.m. - 12:00 p.m
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      Getting 'large enough' data sets is a problem for most deep learning applications, and this is particularly true in medical imaging. Generative Adversarial Networks (GANs) are a deep learning technology in which a computer is trained to create images that look very 'real' even though they are completely synthetic. This may be one way to address the 'data shortage' problem in medicine.
  •  
  • 11:45 a.m. - 1:15 p.m.
    NVIDIA Booth #6568
    Demo

       
    •  
       
      Imagia will demo a clinician driven end-to-end AI biomarker discovery.
  •  
  • 12:30 p.m. - 2:00 p.m.
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This session will focus on multimodal classification. Classification is the recognition of an image or some portion of an image being of one type or another, such as 'tumor' or 'infection'. Multimodal classification means that there are more than 2 classes. While this is logically simple to understand, it presents some unique challenges that will be discussed.
  •  
  • 1:30 p.m. - 3:00 p.m.
    NVIDIA Booth #6568
    Demo

       
    •  
       
      Subtle Medical will demo SubtlePETTM, an artificial intelligence (AI)-powered technology that enables centers to deliver a faster and safer patient scanning experience, while enhancing exam throughput and provider profitability. The new technology is aimed at supporting a four-fold acceleration of scan time or a radiation dose reduction of up to 75%.
  •  
  • 2:30 p.m. - 4:00 p.m.
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This class will focus on basic concepts of convolutional neural networks (CNNs), and walk the attendee through a working example. A popular training example is the MNIST data set which consists of hand-written digits. This course will use a data set we created, that we call 'MedNIST' and consists of 1000 images each from 5 different categories: Chest X-ray, hand X-ray, Head CT, Chest CT, Abdomen CT, and Breast MRI. The task is to identify the image type. This will be used to train attendees on the basic principles and some pitfalls in training a CNN. The attendee will have the best experience if they are familiar with Python programming.
  •  
  • 3:15 p.m. - 4:45 p.m.
    NVIDIA Booth #6568
    Demo

       
    •  
       
      HeartVista's work centers on developing real-time MRI applications for the diagnosis of cardiovascular disease..

TUESDAY 11/27

  • 8:30 a.m. - 10:30 a.m.
    Deep Learning & Machine Intelligence in Radiology
    Abdul Hamid Halabi, Global Business Development Lead, Healthcare and Life Sciences(NVIDIA)

    Room S406A
    Talk
    Share

       
    •  
       
      Current and near future requirements and constraints will require radiology practices to continuously improve and demonstrate the value they add to the healthcare enterprise. Merely 'managing the practice' will not be sufficient; groups will be required to compete in an environment where the goal will be measurable improvements in efficiency, productivity, quality, and safety. There has been great interest (as well as fear and hype) regarding the application of deep learning and other machine intelligence approaches to help improve the radiology value proposition. This session will attempt to provide a ""reality check"" on how these potentially promising technologies might be used by radiology and the significant challenges involved. Topics that will be covered include: How can we best apply deep learning/machine intelligence to add ""true value?"" How do we confidently validate the performance of these technologies? How can our existing IT systems ""feed and consume"" these technologies efficiently and at scale? How can we best harmonize the human radiologist with these machine agents?
  •  
  • 8:30 a.m. - 10:00 p.m.
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This session will focus on the use of deep learning methods for segmentation, with particular emphasis on 3D techniques (V-Nets) applied to the challenge of MR brain segmentation. While focused on this particular problem, the concepts should generalize to other organs and image types.
  •  
  • 10:00 a.m. - 11:30 p.m.
    NVIDIA Booth #6568
    Demo

       
    •  
       
      Ne Scientific focuses on using GPUs to provide real-time simulations during thermal ablation of tumors providing surgical guidance solutions to surgeons in the operating room.
  •  
  • 10:30 a.m. - 12:00 p.m
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This class will focus on basic concepts of convolutional neural networks (CNNs), and walk the attendee through a working example. A popular training example is the MNIST data set which consists of hand-written digits. This course will use a data set we created, that we call 'MedNIST' and consists of 1000 images each from 5 different categories: Chest X-ray, hand X-ray, Head CT, Chest CT, Abdomen CT, and Breast MRI. The task is to identify the image type. This will be used to train attendees on the basic principles and some pitfalls in training a CNN. The attendee will have the best experience if they are familiar with Python programming.
  •  
  • 11:45 a.m. - 1:15 p.m.
    NVIDIA Booth #6568
    Demo

       
    •  
       
      Imagia will demo a clinician driven end-to-end AI biomarker discovery.
  •  
  • 12:30 p.m. - 2:00 p.m.
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This session will focus on preparation of the image and non-image data in order to obtain the best results from your deep learning system. It will include a discussion of different options for representing the data, how to normalize the data, particularly image data, the various options for image annotation and the benefits of each option. We will also discuss the 'after training' aspects of deep learning including validation and testing to ensure that the results are robust and reliable.
  •  
  • 1:30 p.m. - 3:00 p.m.
    NVIDIA Booth #6568
    Demo

       
    •  
       
      EMTensor will explain the physical principles of electromagnetic tomography and the process to reconstruct an image of the brain and present results from their ongoing clinical studies with stroke patients.
  •  
  • 2:30 p.m. - 4:00 p.m.
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This session will focus on multimodal classification. Classification is the recognition of an image or some portion of an image being of one type or another, such as 'tumor' or 'infection'. Multimodal classification means that there are more than 2 classes. While this is logically simple to understand, it presents some unique challenges that will be discussed.
  •  
  • 4:30 p.m. - 6:00 p.m.
    Deep Learning - An Imaging Roadmap
    Abdul Hamid Halabi, Global Business Development Lead, Healthcare and Life Sciences(NVIDIA)
    Room E451B
    Talk
    Share

       
    •  
       
      This overview session of Deep Learning will provide a clearer picture by presenters who are active in that field and who can clarify how the unique characteristics of Deep Learning could impact clinical radiology. It will address how radiologists can contribute to, and benefit from, this new technology. Topics of this multi-speaker session will cover: 1) the general principles of deep learning computational schemas and their mechanisms of handling image inputs and outputs. 2) new technology including hardware shifts in microprocessors from CPU's to GPU devices that offer significant computational advantages 3) how to ensure that Deep Learning results are consistently clinically relevant and meaningful including nodal element tuning and provability so as to assure medical care consistency and reproducibility. 4) how to develop and leverage datasets for deep learning on archives such as the NIH The Cancer Imaging Archive (TCIA) including requirements for input image dataset magnitude and completeness of disease spectrum representation. 5) how to embed essential non-imaging data needed as inputs, (e.g. EHR, outcome, cross-disciplinary metadata, and the data pre-processing required to make DICOM ready for Deep Learning. The presentations will be at a level understandable and relevant to the RSNA radiologist audience.
  •  
  • 3:15 p.m. - 4:45 p.m.
    NVIDIA Booth #6568
    Demo

       
    •  
       
      This demo will focus on ImageBiopsy Lab's technology to advance medical imaging.

WEDNESDAY 11/28

  • 8:30 a.m. - 10:00 p.m.
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This session will focus on preparation of the image and non-image data in order to obtain the best results from your deep learning system. It will include a discussion of different options for representing the data, how to normalize the data, particularly image data, the various options for image annotation and the benefits of each option. We will also discuss the 'after training' aspects of deep learning including validation and testing to ensure that the results are robust and reliable.
  •  
  • 10:00 a.m. - 11:30 p.m.
    NVIDIA Booth #6568
    Demo

       
    •  
       
      This demo will focus on two main algorithms, anatomy segmentation and tracking-less 3D reconstruction on carotid US scans.
  •  
  • 10:30 a.m. - 12:00 p.m
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This class will focus on basic concepts of convolutional neural networks (CNNs), and walk the attendee through a working example. A popular training example is the MNIST data set which consists of hand-written digits. This course will use a data set we created, that we call 'MedNIST' and consists of 1000 images each from 5 different categories: Chest X-ray, hand X-ray, Head CT, Chest CT, Abdomen CT, and Breast MRI. The task is to identify the image type. This will be used to train attendees on the basic principles and some pitfalls in training a CNN. The attendee will have the best experience if they are familiar with Python programming.
  •  
  • 11:45 a.m. - 1:15 p.m.
    NVIDIA Booth #6568
    Demo

       
    •  
       
      Imagia will demo a clinician driven end-to-end AI biomarker discovery.
  •  
  • 12:30 p.m. - 2:00 p.m.
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This class will focus on basic concepts of convolutional neural networks (CNNs), and walk the attendee through a working example. A popular training example is the MNIST data set which consists of hand-written digits. This course will use a data set we created, that we call 'MedNIST' and consists of 1000 images each from 5 different categories: Chest X-ray, hand X-ray, Head CT, Chest CT, Abdomen CT, and Breast MRI. The task is to identify the image type. This will be used to train attendees on the basic principles and some pitfalls in training a CNN. The attendee will have the best experience if they are familiar with Python programming.
  •  
  • 1:30 p.m. - 3:00 p.m.
    NVIDIA Booth #6568
    Demo

       
    •  
       
      CureMetrix will demo a physics-based AI solution and how they can be implemented into your breast screening workflow for earlier and more accurate detection of breast cancer.
  •  
  • 2:30 p.m. - 4:00 p.m.
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This session will focus on the use of deep learning methods for segmentation, with particular emphasis on 3D techniques (V-Nets) applied to the challenge of MR brain segmentation. While focused on this particular problem, the concepts should generalize to other organs and image types.
  •  
  • 3:15 p.m. - 4:45 p.m.
    NVIDIA Booth #6568
    Demo

       
    •  
       
      This demo will focus on a Clara enabled lumbar spinal stenosis classification workflow. Radiologists will be able to load lumbar scan data from a patient and at the click of a button, will send the patient images for evaluation by AI. The results will then be available for download at the originating dicom server under the same patient id, providing an opportunity to evaluate the results of the AI.

THURSDAY 11/29

  • 8:30 a.m. - 10:00 p.m.
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This session will focus on the use of deep learning methods for segmentation, with particular emphasis on 3D techniques (V-Nets) applied to the challenge of MR brain segmentation. While focused on this particular problem, the concepts should generalize to other organs and image types.
  •  
  • 10:00 a.m. - 11:30 p.m.
    NVIDIA Booth #6568
    Demo

       
    •  
       
      This demo will focus on showcasing inference with GPUs and talking about leveraging multiple models from several companies to deploy in clinic.
  •  
  • 10:30 a.m. - 12:00 p.m
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This session will focus on multimodal classification. Classification is the recognition of an image or some portion of an image being of one type or another, such as 'tumor' or 'infection'. Multimodal classification means that there are more than 2 classes. While this is logically simple to understand, it presents some unique challenges that will be discussed.
  •  
  • 11:45 a.m. - 1:15 p.m.
    NVIDIA Booth #6568
    Demo

       
    •  
       
      Imagia will demo a clinician driven end-to-end AI biomarker discovery.
  •  
  • 12:30 p.m. - 1:00 p.m.
    Towards Intelligent Healthcare
    Kimberly Powell,Vice President, Healthcare(NVIDIA)
    Machine Learning Pavillion Theatre
    Talk
    Share

       
    •  
       
      With the rise of artificial intelligence (AI), computing has the potential to go far beyond imaging technology--reinventing radiology workflows and greatly improving the hospital's ability to serve patients. The NVIDIA GPU platform is enabling the entire medical imaging ecosystem from researchers and radiologists to providers and healthcare IT to startups. This session discusses how NVIDIA is evolving the AI computing platform to rapidly build and deploy deep learning applications in radiology.
  •  
  • 12:30 p.m. - 2:00 p.m.
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      Getting 'large enough' data sets is a problem for most deep learning applications, and this is particularly true in medical imaging. Generative Adversarial Networks (GANs) are a deep learning technology in which a computer is trained to create images that look very 'real' even though they are completely synthetic. This may be one way to address the 'data shortage' problem in medicine.
  •  
  • 2:30 p.m. - 4:00 p.m.
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This session will focus on preparation of the image and non-image data in order to obtain the best results from your deep learning system. It will include a discussion of different options for representing the data, how to normalize the data, particularly image data, the various options for image annotation and the benefits of each option. We will also discuss the 'after training' aspects of deep learning including validation and testing to ensure that the results are robust and reliable.

FRIDAY 11/30

  • 8:30 a.m. - 10:00 p.m.
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      This session will focus on multimodal classification. Classification is the recognition of an image or some portion of an image being of one type or another, such as 'tumor' or 'infection'. Multimodal classification means that there are more than 2 classes. While this is logically simple to understand, it presents some unique challenges that will be discussed.
  •  
  • 10:30 a.m. - 12:00 p.m
    Learning Center (Hall D)
    Hands On Training

       
    •  
       
      Getting 'large enough' data sets is a problem for most deep learning applications, and this is particularly true in medical imaging. Generative Adversarial Networks (GANs) are a deep learning technology in which a computer is trained to create images that look very 'real' even though they are completely synthetic. This may be one way to address the 'data shortage' problem in medicine.